3. Weather Data Driven Analysis of Renewable Energy Generation 31
3.2. Data Time Series
Key to the modeling of wind- and solar-power generation is a large weather data set with good spatial and temporal resolution all over Europe. Its convolution with future-projected wind and solar power capacities reveals how much wind and solar power is generated across Europe together with the spatial and temporal information. The following subsections will explain the details. The load modeling is described in the last subsection of this section.
3.2.1. Weather Data
Weather data for all of Europe is available from various sources with different spatial and temporal resolutions. For our purposes three selection criteria have been important:
• The correct modeling of intra-day solar- and wind-power ramps require a good
3.2. Data Time Series
Figure 3.6.: The region partintioning. Due to the fact that load data is only available for regions larger than the grid size of the weather data, the generation data is aggregated accordingly. The gray line indicates the onshore area, regions outside are considered as offshore. The borders between the regions are indicated by the black lines.
time resolution of at least 1h.
• In order to resolve the passing of synoptic systems related to high winds and opaque clouds a spatial resolution of at least 50×50km2 is required.
• In order to gain representative and significant statistics covering all possible seasonal and extreme weather situations a rather long time window is required, ranging over a couple of years.
These criteria have been met by the private weather service provider WEPROG (Weather & Wind Energy Prognosis) [124]. With regional models it downscales medium-resolved analysis data from the US Weather Service NCEP (National Center for Envi-ronmental Prediction) [86] down to 47×48km2 spatial and 1h time resolution over an eight-years period (2000-2007).
This high-resolution data provides direct information on the wind speed and direction 100m above ground. The solar global radiation is not a standard output, but can be computed directly from the data on the net short wave radiation at the surface, the
0.00.10.20.30.40.50.60.7
(a)Wind power capacities
0.00.10.20.30.40.5
(b) Solar photovoltaics power capacities Figure 3.7.: Expected wind power and solar photovoltaics power capacities [GW] per grid-cell
across Europe in 2020. The spatial grid-cell resolution of 47×48km2 has been adapted to the weather data. For a better visualization capacities larger than 0.73 GW for wind and 0.50 GW for PV are indicated in dark red.
total cloud cover, and a standard cloud and surface albedo.
3.2.2. Assumed Wind- and Solar-Power Capacities and Generation
The national 2020 targets serve as guidance for a rough distribution of wind and photovoltaic capacities in Europe. Figure 3.7 illustrates the expected installed wind-power and solar photovoltaics wind-power capacities across Europe. They total to 227 and 68 GW, respectively. 66 GW of wind power is assumed to be installed offshore. The subsequent finer distribution within each country onto the grid cells of the weather data is done empirically, giving more capacity to those grid cells with large average wind speed and large average global radiation, respectively.
The conversion of hourly WEPROG wind speeds into wind power at each grid cell was done using typical wind power curves at 100m hub height. Different power curves have been assigned for on- and offshore grid points. Losses due to wake effects have been modeled explicitly for offshore grid points by assuming a park layout of 7x7 turbines in offshore wind farms. Additional 7% losses have been introduced due to electrical losses and turbine non-availability. The same 7% of losses have also been applied to onshore grid points. The turbine cut-off due to extreme winds is empirically parameterized by an additional modification of the power curve, which mimics the gradual power-lowering-behavior of wind turbines with storm-control.
The solar photovoltaic power generation within the grid cells has been calculated based
3.2. Data Time Series
0 20 40 60 80
0.850.951.051.15
t [month]
normalzed power
Figure 3.8.: The monthly averaged load accumulated over Europe. The black curve indicates the data. An exponential trend is removed resulting in the load shown by the red curve. The black dashed line shows the fit of an exponential trend based on a running one year average.
on the available meteorological data (global radiation, air temperature), assumptions on the characteristics of the photovoltaic plants (tilt angle, orientation, fixed or with solar tracker) and the geographical coordinate of the grid cell considered. A mix of different photovoltaic plant technologies was considered for each grid cell [16].
This convolution of the weather data with the wind and solar power capacities produces spatio-temporal power-generation patterns across Europe. These patterns are important for the calculation of power flows, see Chapter 6.
3.2.3. Load modeling
There are no sources to obtain the load data on the same spatial 47×48km2 grid-cell resolution as for generation. But for the work presented here, a coarser resolution is sufficicent. For almost all European countries the load profiles have been downloaded either from the UCTE-homepage [117] or from the national transmission providers. At least for the two recent years those have an hourly resolution. For the remaining years they have been replicated with the known relative annual electric power consumption;
special care was given to a proper handling of the weekend effect.
The load increases over time, see Figure 3.8. Since for generation the fixed capacities expected for 2020 are used, the load is detrended. An exponential trend2 is removed according to y=aexp(b·t). Every country is detrended individually, since the growth factor strongly deviates from country to country. For all of Europe the load increases
2The exponential detrending was not done in Heide et al. [64], so the results shown here differ slightly.
1020304050
Figure 3.9.: Average annual load [TWh] per grid cell in the 50 coarse-grained onshore regions.
by 1.138% per year over the considered eight years.
Some countries, especially the larger ones, come with a large average load. Those have been further subdivided into regions, with some spatial correlation to the territories of the respective network transmission providers. The regional load profiles have been obtained from the country profiles with a multiplicative factor obtained from a linear regression between the annual electric power consumption on the one hand and population and gross domestic product on the other hand.
Figure 3.9 shows the average annual load of the 50 onshore regions during the years 2000-2007. Offshore regions come with no load and are not shown. The sum over all regions totals to 2995 TWh average annual consumption. Its seasonal dependence is shown in Figure 3.10.